Worms: Identifying Impacts on Education and Health in the Presence of Treatment Externalities

Publication Details

Econometrica, January 2004, vol.72, iss.1, pp.159-217. Available From:

Link to Source
Edward Miguel, Michael Kremer
Institutional affiliations
None specified
Grant-holding institution
None specified
Sub-Saharan Africa (includes East and West Africa)
Education, Health Nutrition and Population
Pre-Primary and Primary Education, Child Nutrition, Health Services, Preventive Health and Health Behavior
Pre-Primary and Primary Education, Child Nutrition, Health Services, Preventive Health and Health Behavior
Equity Focus
Evaluation design
Randomised Control Trials (RCT)
Journal Article


This study evaluates the impact of a deworming programme in Kenya on health and education outcomes. It highlights the importance of accounting for positive treatment externalities, as failing to do so leads evaluators to severely underestimate the impact. By improving overall health, the mass treatment of children aims to increase their participation and performance in school. Overall lower infection prevalence also benefits untreated children in school and in surrounding communities. The authors estimate these positive externalities to consider them in their final cost-benefit analysis.

The sample includes 75 primary schools in a rural region of Kenya. The project schools were randomly divided into three groups of 25 schools each. Group 1 received treatment in 1998, whilst groups 2 and 3 functioned as a control group in the first year of programme implementation. In 1999, group 2 schools were also treated, reducing the control group to group 3. Pupil and school surveys were conducted in early 1998 and 1999 (right before group 2 began treatment).

Treatment is administered to all pupils at one school, thereby randomising treatment at the school rather than individual level. This feature allows estimation of the overall impact of treatment on the school, including within-school treatment externalities. To decompose the treatment effect into direct and externality effects, the authors identify children who did not comply with treatment (and benefitted through externality effects) in group 1 and those who self-selected out of treatment in group 2, 1 year later. They then compare infection rates between these two groups in early 1999 (before group 2 started treatment).

Finally, to estimate cross-school externalities, the authors evaluate the impact of a higher density of treated pupils on children within 3 km of treatment schools. By carefully defining treatment and control groups for each kind of treatment effect, the authors can conduct simple regressions to measure the average treatment effect.

Main findings

The authors first present the differences in health outcomes between treatment and control schools in early 1999: whereas 52 per cent of pupils in group 2 (1998 control) had moderate to heavy worm infections, the rate was only 27 per cent in treatment schools. In early 1999, infection rates amongst pupils in group 1 who did not receive treatment in 1998 were 21 percentage points lower than amongst group 2 pupils who selected out of treatment during 1999. This difference, which serves as a proxy to estimate within-school externalities, is significant at the 5 per cent level.

Finally, the authors find that the presence of each additional thousand pupils from treatment schools, located within 3 km of another primary school, is associated with 26 percentage points fewer infections. This difference is significant at the 1 per cent level. Through its significant impact on health outcomes, deworming has further increased school participation in treatment schools by at least 7 percentage points. Nevertheless, the authors could not identify any significant impact of deworming on test scores.

A cost-benefit analysis of the deworming programme concludes that the substantial positive spillovers are sufficient to justify fully subsidising deworming treatment and perhaps even paying people to receive treatment. Future research should foresee a multilevel design that randomises treatment within and across schools and amongst school clusters to learn even more about the magnitude of externalities. In view of these encouraging findings, the authors provide a rationale for a substantial public role in the fight against infectious diseases.

Scroll to Top